Re: [Xen-devel] BalloonWorkerThread issue

[Paul Durrant <Paul.Durrant@xxxxxxxxxx>]

--
From: R J [mailto:torushikeshj@xxxxxxxxx] 
Sent: 06 January 2012 21:13
To: Paul Durrant
Cc: Konrad Rzeszutek Wilk; annie.li@xxxxxxxxxx; Konrad Rzeszutek Wilk; 
xen-devel@xxxxxxxxxxxxxxxxxxx; xen-users@xxxxxxxxxxxxxxxxxxx; 
xen-api@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-devel] BalloonWorkerThread issue

Hello Paul,

Thanks for your email and explanation.  

I'm trying multiple combinations and found that windows will run stable only if 
the static max is twice bigger.
I did the test on below static min and max

512MB to 2GB   <- Pass
1GB to 4 GB    <-  Some times pass, some time fails
2GB to 4GB <- Pass
2GB to 8 GB <- Pass
4GB to 32 GB <-- Fail
16GB to 32GB <-- Pass


So it seems that the problem is not due to size of RAM but its due to 
difference between them.
Is there any defined multiplication factor while initial squeeze down ?

Interesting thing is if I start a VM with Static max 32GB and dynamic max, 
dynamic min 32GB and static min 512MB then it starts fine and is able to boot 
successfully. The reason here is no ballooning required as target is equal to 
static max.

Once the VM is up and if I set its memory target to 1 GB ( squeezing from 32G 
to 1G)  it works fine. No issue of balloon driver or anything.
So I did same for other cases as well where static max and target were same. 
The result was "pass".

Its only the boot process which is hampered.
--

I believe top-posting is against etiquette for this list so I won't continue 
it...

I don't think anyone has ever determined a multiplication factor that will 
cover *any* windows sku... there's too much variation between them. It's not 
that surprising that ballooning down after boot gives better results since 
booting will almost certainly require more code and data to be paged in.

  Paul

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel